Monochromatic subgraphs in iterated triangulations
Extracted source text · arXiv 1912.00123
\documentclass[11pt]{article}
\usepackage{amssymb}
\usepackage{mathrsfs}
\addtolength{\topmargin}{-.5in} \addtolength{\textheight}{1in}
\addtolength{\oddsidemargin}{-.5in}
\addtolength{\evensidemargin}{-.5in}
\addtolength{\textwidth}{1in}
\usepackage{latexsym,amsmath,amssymb,amsfonts,epsfig,graphicx,cite,psfrag}
\usepackage{eepic,color,colordvi,amscd}
\usepackage{ebezier}
\usepackage{verbatim}
\usepackage{subfigure}
\usepackage{enumitem}
\usepackage{amsthm}
\usepackage{comment}
\usepackage{color}
\theoremstyle{plain}
\newtheorem{theo}{Theorem}[section]
\newtheorem{claim}[theo]{Claim}
\newtheorem{ques}[theo]{Question}
\newtheorem{lem}[theo]{Lemma}
\newtheorem{coro}[theo]{Corollary}
\newtheorem{cors}[theo]{Corollaries}
\newtheorem{prop}[theo]{Proposition}
\newtheorem{conj}[theo]{Conjecture}
%\newtheorem{defi}[theo]{Definition}
\newtheorem{nota}[theo]{Notation}
\newtheorem{exmp}[theo]{Example}
\newtheorem{prob}[theo]{Problem}
%\newtheorem{rem}[theo]{Remark}
\newtheorem{ack}{Acknowledgments}
\theoremstyle{definition}
\newtheorem{defi}[theo]{Definition}
\newtheorem*{goal}{Goal}
\theoremstyle{remark}
\newtheorem{rem}[theo]{Remark}
\setcounter{section}{0}
\def\qed{\hfill \rule{4pt}{7pt}}
\def\pf{\noindent {\it Proof.\quad}}
\def\Tr{\mathrm{Tr}}
\let\svthefootnote\thefootnote
\newcommand\blankfootnote[1]{%
\let\thefootnote\relax\footnotetext{#1}%
\let\thefootnote\svthefootnote%
}
\pagestyle{plain}
\begin{document}
\title{Monochromatic subgraphs in iterated triangulations
%\footnote{Partially supported by NSF grant DMS-1600738.
%Email: qxie7@gatech.edu (Q. Xie), shijie.xie@gatech.edu (S. Xie), yu@math.gatech.edu (X. Yu), and xyuan@gatech.edu (X. Yuan)}
}
\author{Jie Ma\footnote{Email: jiema@ustc.edu.cn. Partially supported by NSFC grant 11622110.} ~ and ~Tianyun Tang\footnote{Email: tty123@mail.ustc.edu.cn.}\\ School of Mathematical Sciences\\University of Science and Technology of China\\Hefei, Anhui 230026, China\\
\medskip\\
Xingxing Yu\footnote{Email: yu@math.gatech.edu. Partially supported by NSF grant DMS-1600738.} \\
School of Mathematics\\
Georgia Institute of Technology\\
Atlanta, GA 30332, USA}
\date{}
\maketitle
\begin{abstract}
%A graph $H$ is said to be {\it planar unavoidable} if there is a planar graph $G$ such
%that for any 2-edge-coloring of $G$, there is a monochromatic copy of $H$. We denote this by $G\to H$. In \cite{ASTU19} the authors
%observed that planar unavoidable graphs must be bipartite and outer planar. They also asked whether even cycles of length at least 6 are planar
%unavoidable, and whether for every planar unavoidable graph $H$ there exists a special triangulation $\Tr(n)$ such that $\Tr(n)\to H$.
For integers $n\ge 0$, an iterated triangulation $\Tr(n)$ is defined recursively as follows:
$\Tr(0)$ is the plane triangulation on three vertices and, for $n\ge 1$, $\Tr(n)$ is the plane triangulation
obtained from the plane triangulation $\Tr(n-1)$ by, for each inner face $F$ of $\Tr(n-1)$,
adding inside $F$ a new vertex and three edges joining this new vertex to the three vertices incident with $F$.
In this paper, we show that there exists a 2-edge-coloring of $\Tr(n)$ such that $\Tr(n)$ contains no monochromatic copy of the cycle $C_k$ for any $k\ge 5$. As a consequence,
the answer to one of two questions asked in \cite{ASTU19} is negative.
%$\Tr(n)\not \to C_k$ for any $k\ge 5$ which implies that the answer to one of these two questions is negative.
We also determine the radius two graphs $H$ for which there exists $n$ such that every 2-edge-coloring of $\Tr(n)$
contains a monochromatic copy of $H$, extending a result in \cite{ASTU19} for radius two trees.
\blankfootnote{AMS Subject Classification: 05C55, 05C10, 05D10}
\blankfootnote{Keywords: Triangulation, Planar unavoidable, Coloring, Monochromatic subgraph}
\end{abstract}
\section{Introduction}
For graphs $G$ and $H$, we write $G \to H$ if, for any 2-edge-coloring of $G$, there is a monochromatic
copy of $H$. Otherwise, we write $G \not\to H$.
We say that $H$ is {\it planar unavoidable} if there
exists a planar graph $G$ such that $G\to H$. This notion is introduced and studied in \cite{ASTU19}.
Deciding if $G\to H$ is clearly equivalent to asking whether a graph $G$ admits a decomposition (i.e., an edge-decomposition)
such that one of the two graphs in the decomposition contains the given graph $H$.
The well-known Four Color Theorem \cite{AH77, AHK77} (also see \cite{RSST98}) implies that every planar graph admits
a decomposition to two bipartite graphs; so planar unavoidable graphs must be bipartite.
A result of Goncalves \cite{Go05} says that every planar graph admits a decomposition to two outer planar graph (although we have not seen a detailed proof); so planar unavoidable graphs must be also outer planar.
There are a number of interesting results about decomposing planar graphs, see \cite{AA89, CCW86, Ha65, HMS96, Po90}.
For any positive integer $n$, let $P_n$ denote the path on $n$ vertices, and $K_n$ denote the complete graph on
$n$ vertices. For integer $n\ge 3$, we use $C_n$ to denote the cycle on $n$ vertices.
It is shown in \cite{ASTU19} that $P_n$, $C_4$, and all trees with radius at most 2 are planar unavoidable.
This is done by analyzing several sequences of graphs.
In this paper, we investigate one such sequence -- the iterated triangulations,
which is of particular interest as suggested in \cite{ASTU19}.
Let $n\ge 0$ be an integer.
An {\it iterated triangulation} $\Tr(n)$ is a plane graph defined as follows:
$\Tr(0)\cong K_{3}$ is the plane triangulation with exactly two faces.
For each $i\ge 0$, let $\Tr(i+1)$ be obtained from the plane triangulation $\Tr(i)$
by adding a new vertex in each of the inner faces of $\Tr(i)$ and connecting this vertex with edges to the three vertices in the boundary of their respective face.
% (See Figure 1)
%\begin{figure}
% \centering
% \includegraphics[scale=0.5]{001.png}
% \caption{The iterative construction of Tr(2)}
%\end{figure}
%See Figure 1. We know that \Tr(i) is a triangulation and all the inner faces of \Tr(i) are triangles.
The authors of \cite{ASTU19} asked whether for any planar unavoidable graph $H$ there is an integer $n$ such
that $\Tr(n)\to H$.
They also asked whether there exists an integer $k\ge 3$ such that the even cycle $C_{2k}$ is planar-unavoidable.
Our first result indicates that a positive answer to one of the above questions implies a negative answer to the other.
Let $H^+$ be the bipartite graph obtained by adding an edge to the unique 6-vertex tree with 4 leaves and 2 vertices of degree three.
%Specifically, we show that there exists a 2-edge-coloring of $\Tr(n)$ resulting
%in no monochromatic $C_{k}$ for any $k\geq 5$, no monochromatic $K_{2,3}$, and no monochromtic $H^+$.
\begin{theo}\label{main}
For all positive integer $n$, $\Tr(n) \not\to C_{k}$ for $k \geq 5$, $\Tr(n) \not\to H^+$, and $\Tr(n) \not\to K_{2,3}$
\end{theo}
As another direct consequence, we see that if $B$ is a bipartite graph and $\Tr(n)\to B$ for some $n$ then every block of $B$ must be a $C_4$ or $K_2$. This can be used to
characterize all radius two graphs $B$ for which there exists $n$ such that $\Tr(n)\to B$,
generalizing a result in \cite{ASTU19} for radius two trees.
To state this characterization, we need additional notations.
A {\it flower} $F_{k}$ is a collection of $k$ copies of $C_4$s sharing a common vertex, which is called the {\it center}.
%A {\it flower} $F_{k}$ is obtained from $k$ pairwise disjoint $C_4$s by taking one vertex from each $C_4$ and identifying them as a
%single vertex, called the {\it center} of $F_k$.
A {\it jellyfish} $J_{k}$ is obtained from $F_k$ and a $k$-ary tree of radius two by
identifying the center of $F_{k}$ with the root of the $k$-ary tree.
A {\it bistar} $B_k$ is obtained from one $C_4$ and two disjoint $K_{1,k}$s by
identifying the roots of the $K_{1,k}$s with two non-adjacent vertices of $C_4$, respectively.
\begin{theo}\label{radius2}
Let $L$ be a graph with radius two. Then there exists $n$ such that $\Tr(n)\to L$ if, and only if, $L$ is a subgraph of a jellyfish or bistar.
\end{theo}
We organize this paper as follows. In Section 2, we prove $\Tr(n) \not\to C_{k}$ for $k \geq 5$ and $\Tr(n) \not\to H^+$ by finding a special edge coloring scheme for $\Tr(n)$.
In Section 3, we complete the proof of Theorem~\ref{main} by using another edge coloring scheme on $\Tr(n)$.
From Theorem \ref{main}, we can derive the following:
if $L$ has radius 2 and $\Tr(n)\to L$ for some $n$, then $L$ is a subgraph of a jellyfish or bistar.
Hence to prove Theorem~\ref{radius2}, it suffices to show that for any $k\ge 1$ there exists some $n$ such that $\Tr(n)\to J_k$ and $\Tr(n)\to B_k$.
We prove the former statement in Section 4 and the latter one in Section 5 by showing that we can choose $n$ to be linear in $k$.
\begin{comment}
\begin{lemma}
For any positive integer $k$, $\Tr(100k)\to J_k$ and $\Tr(6k+30)\to B_k$.
\end{lemma}
Theorem~\ref{main} implies that, for any planar unavoidable graph $H$, every block of $H$ must
be a $C_4$ or $K_2$. Hence, it is interesting to know when such graphs are are planar unavoidable.
We show that certain radius 2 graphs whose blocks are $C_4$ or $K_2$ are planar unavoidable,
generalizuing the result in ~\cite{ASTU19} that trees of radius 2 are planar unavoidable.
\begin{theo}\label{ax}
Let $H$ be the connected graph such that $H\cong K_{2,3}$ or $H$ consists of three blocks, a $C_4$ and two disjoint $K_2$s
Then for $n>0$, $\Tr(n) \nrightarrow H$.
\end{theo}
\end{comment}
%First, we find a new iterative coloring method of $\Tr(n)$, which avoids another simple graph.
%\begin{theo}\label{new}
%$G$ is graph $s.t$ $V(G)=\{v_1,v_2,v_3,v_4,v_5\}$, $\{v_1,v_2,v_3,v_4\}$
%induces a $C_4$ $v_1v_2v_3v_4v_1$ and $N_G(v_5)$=$\{v_2,v_4\}$. Then
% $\Tr(n)\nrightarrow K_{2,3}$ for $\forall n \in N^+$.
%\end{theo}
%\begin{figure}
% \centering
% \includegraphics[scale=0.7]{15.png}
% \caption{$F_4$, $J_4$, $B_4$}
%\end{figure}
%In addition, we combine several $C_4$s and a $k-ary-perfect$ tree with radius 2 to form a new graph, which
%seems like a jellyfish.
\begin{comment}
\begin{theo}\label{jelly}
For any positive integer $k$, $\Tr(100k)\toJ_k$.
\end{theo}
2, Suppose $K_1,K_2$ are two $K_{1,k+2}$s and $u_1,v_1$ are $K_1$'s two leaves.
$u_2,v_2$ are $K_2$'s two leaves. A bistar $B_k$ is obtained by identifying $u_1\
with\ u_2$ and identifying $v_1\ with\ v_2$, see Figure 2.
%We will proof $J_k_x$ is planar-unavoidable.
\begin{theo}\label{bistar}
For any positive integer $k$, $\Tr(6k+30)\to B_k$.
\end{theo}
%\begin{lem}\label{5cuttri}
% ......
%\end{lem}
\section{Definitions}
We also use abca to denote a $K_{3}$ with vertices a,b,c which appears
anticlockwitely. We also use some definitions in \cite{A.M} directly.
\begin{enumerate}
\item Iterated Triangulation \Tr(n)\cite{A.M}:\\
An iterated triangulation is a planar graph \Tr(n) defined as follows: \Tr(0)=$K_{3}$ is a triangle, \Tr(i+1) is obtained from \Tr(i) by inserting a vertex in each of the inner faces of \Tr(i) and connecting this vertex with edges to all the boundary of the respective face. See Figure 1. We know that \Tr(i) is a
triangulation and all the inner faces of \Tr(i) are triangles.
% \begin{figure}
% \centering
% \includegraphics[scale=0.5]{001.png}
% \caption{The iterative construction of Tr(2)}
% \end{figure}
\item Center inner point:\\
o is the center inner point of abca in \Tr(i) for some i if o appears in abca
after a new iteration i.e o is connected to a,b and c.
\item$2_{nd}$ center inner point:\\
$2_{nd}$ center point $o_{1},o_{2},o_{3}$ of abca are those points appear in abca
after two new iterations. In that, there are 3 $2_{nd}$ center inner point of
abca. See Figure1, o is the center inner point of abca. o1, o2, o3 are the
$2_{nd}$ center inner point of abca.
\item Y:\\
o is the center inner point of abca. Then we use $Y_{(o)abc}$ to denote the
subgraph with points o, a, b, c and edges oa, ob, oc.
\item Fish\cite{A.M}:A graph G is called a fish and denote $F_{x,y}$ if
V(G)=\{x,y\}$\cup$S, where S$\cap$\{x,y\}=$\phi$, x and y are adjacent to each
vertex in S. S itself induces a path in G, and xy is an edge. We call S the set
of spine vertices and G[S] the spine. We call xs, ys ribs for each s $\in$ S(see Figure 2).
% \begin{figure}
% \centering
% \includegraphics[scale=0.4]{002.png}
% \caption{A fish $F_{x,y}$ with k spine vertices}
% \end{figure}
\item Coloring function CR:\\
CR: E(\Tr(n)) $\rightarrow$ \{0,1\}. CR(e)=0 iff e is red. CR(e)=1 iff e is green.
\end{enumerate}
\end{comment}
\section{$H^+$ and $C_k$ for $k\geq 5$}
In this section, we prove Theorem~\ref{main} for $H^+$ and $C_k$, with $k\ge 5$. First, we
describe the 2-edge-coloring of $\Tr(n)$ that we will use.
Let $\sigma: E(\Tr(n))\to \{0,1\}$ be defined inductively for all $n\geq 1$ as follows:
%For each $e\in E(\Tr(n))$, let
\begin{itemize}
\item [(i)] Fix an arbitrary triangle $T$ bounding an inner face of $\Tr(1)$, and let $\sigma(e)=0$ if $e\in E(T)$ and $\sigma(e)=1$
if $e\in E(\Tr(1))\setminus E(T)$. %(See Figure 3.)
%\begin{figure}
% \centering
% \includegraphics[scale=0.5]{01.png}
% \caption{First step}
%\end{figure}
\item [(ii)] Suppose for some $1\le i<n$, we have defined $\sigma(e)$ for all $e\in E(\Tr(i))$.
We extend $\sigma$ to $E(\Tr(i+1))$ as following.
Let $x\in V(\Tr(i))\setminus V(\Tr(i-1))$ be arbitrary, let $v_0v_1v_2v_0$ denote the triangle bounding
the inner face of $\Tr(i-1)$ containing $x$, and fix a labeling so that $\sigma(xv_1)=\sigma(xv_2)$.
\item [(iii)] Let $x_j\in V(\Tr(i+1))\setminus V(\Tr(i))$ be such that
$x_j$ is inside the face of $\Tr(i)$ bounded by the triangle $xv_jv_{j+1}x$, where $j=0,1,2$ and the subscripts are taken modulo 3.
Define $\sigma(xv_0)=\sigma(x_0v_0)=\sigma(x_2v_0)=\sigma(x_jx)$ for all $j=0,1,2$, and
$\sigma(xv_1)=\sigma(x_0v_1)=\sigma(x_1v_1)=\sigma(x_1v_2)=\sigma(x_2v_2)$.
% (See Figure 4)
%\begin{figure}
% \centering
% \includegraphics[scale=0.5]{003.png}
% \caption{Second: iterative coloring}
%\end{figure}
\end{itemize}
\begin{comment}
We use coloring function CR to denote the special coloring method of \Tr(n).
Suppose the initial triangle is abca with the center inner point o. \\
First: CR(ab)=CR(ac)=CR(ao)=0, CR(ob)=CR(oc)=CR(bc)=1(see Figure 3).\\
\begin{figure}
\centering
\includegraphics[scale=0.5]{01.png}
\caption{First step}
\end{figure}\\
Second: Suppose o is the center inner point of uvwu. $o_{1}, o_{2}, o_{3}\ are\ the\ 2_{nd}\ center\ inner\ points\ of\ uvwu.\ $ $o_{1}$ is in uvou, $o_{2}$ is in uowu, $o_{3}$ is in ovwo. If $\exists\ v_{1},\ v_{2}\ \in$ \{u,v,w\} s.t CR($v_{1}$o) $\neq$ CR($v_{2}$o), assume CR(ov)=CR(ow)$\neq$CR(ou). \\
Then CR($o_{1}$u)=CR(o$o_{1}$)=CR($o_{2}u$)=CR(o$o_{2}$)=CR(o$o_{3}$)=CR(ou)\\
CR($o_{1}$v)=CR($o_{2}$w)=CR(ov)=CR(ow)=CR($o_{3}$v)=CR($o_{3}$w)(see Figure4).
It is obvious that all Ys in \Tr(n) are not monochromatic according to the method, so
the second rule always works. As a result, every edge of \Tr(n) has a decided color.
\begin{figure}
\centering
\includegraphics[scale=0.5]{003.png}
\caption{Second: iterative coloring}
\end{figure}
\section{Proof of Theorem~\ref{main}}
In this section, we always suppose \Tr(n) is colored by the method mentioned above.\\
Now, we can divide all the edges into three types. An edge e is of type1 if it is an
edge of the initial triangle. e is of type2 if it is in a Y whose color is the same
as another edge in the same Y. e is of type3 in a Y whose color is different from the
other edges in the Y.\\ Our proof has 5 steps.\\
Suppose $o$ is the center inner point of uvwu. If CR(ou)=0, then all edges with one
end point u and the other end point p in uvwu are red. Similarly, If CR(ou)=1, then
all edges with one end point u and the other point p in uvwu are green.\\
Proof by induction:\\
We just prove the first part.\\
Suppose o appears in the $k_{th}$ iteration and p appears in the $k+i_{th}$
iteration(i$\geq$0). If i=0, p is o and CR(up)=CR(ou)=0. (1) satisfies. Suppose (1) is true for i$\leq$r(r$\geq$0). When p appears in the$k+r+1_{th}$iteration. p is $2_{nd}$ center inner point of uefu (both e and f can be v and w) and g is the center inner point of uefu. Without loss of generality, assume p is in ugfu. According to the induction hypothesis, CR(ug)=0. If CR(eg)=CR(gf)=1, then according to the coloring method, CR(up)=CR(gp)=0. CR(pf)=1(see Figure5)\\
If CR(eg)=0, CR(gf)=1, then CR(pf)=CR(pg)=1, CR(up)=0(see Figure5)\\
If CR(eg)=1, CR(gf)=0, then CR(gp)=1, CR(up)=CR(fp)=0(see Figure5)\\
In all the three cases above, (1) satisfies. In that, (1) is proved.\\
\begin{figure}
\centering
\includegraphics[scale=0.5]{004.png}
\caption{proof of (1)}
\end{figure}\\
\end{comment}
%{\it Proof of Theorem~\ref{main}}.
%Let $x\in V(\Tr(i))\setminus V(\Tr(i-1))$ be arbitrary, let $v_0v_1v_2v_0$ denote the triangle bounding
% the inner face of $\Tr(i-1)$ containing $x$, and assume $\sigma(xv_1)=\sigma(xv_2)$ (by relabeling if necessary).
% Let $x_j\in V(\Tr(i+1))\setminus V(\Tr(i))$ such that
% $x_j$ is inside the face of $\Tr(i)$ bounded by the triangle $xv_jv_{j+1}x$, where $j=0,1,2$ and subscripts are taken modulo 3.
We now proceed by a sequence of claims to show that $\sigma$ has no monochromatic $C_k$ for $k\geq 5$ nor monochromatic $H^+$,
thereby proving $\Tr(n)\not\to C_{k}$ for $k\geq 5$ and $\Tr(n)\not\to H^+$.
\begin{itemize}
\item [(1)] For $1\le i\le n$ and $x\in V(\Tr(i))\setminus V(\Tr(i-1))$,
$|\{\sigma(xv): v\in V(\Tr(i-1))\}|=2$.
\end{itemize}
We apply induction on $i$. The basis case $i=1$ follows from (i)
above. So assume $2\le i\le n$.
Let $v_0v_1v_2v_0$ be the triangle bounding the inner face of $\Tr(i-1)$ containing $x$.
Without loss of generality assume that $v_0\in V(\Tr(i-1))\setminus V(\Tr(i-2))$.
Let $v_1v_2v_3v_1$ denote the triangle bounding the face of $\Tr(i-2)$, with $v_0$
inside $v_1v_2v_3v_1$.
By induction hypothesis, $|\{\sigma(v_0v_k): k=1,2,3\}|=2$.
Suppose $\sigma(v_0v_1)=\sigma(v_0v_2)$. Then by (ii) and (iii), $\sigma(xv_0)=\sigma(v_0v_3)$
and $\sigma(xv_1)=\sigma(xv_2)=\sigma(v_0v_1)$. So $|\{\sigma(xv_k): k= 0, 1,2\}|=2$.
So assume $\sigma(v_0v_1)\neq\sigma(v_0v_2)$. By symmetry, we further assume
$\sigma(v_0v_2)=\sigma(v_0v_3)$. Then by (ii) and (iii), we see that $\sigma(xv_1)=\sigma(xv_0)=\sigma(v_0v_1)$
and $\sigma(xv_2)=\sigma(v_0v_2)$.
So $\sigma(xv_2)\ne \sigma(xv_1)$ and hence, $|\{\sigma(xv_k): k=0,1,2\}|=2$. $\Box$
\begin{comment}
\medskip
By (1), we can partition $E(\Tr(n))$ into pairwise disjoint sets $E_0$, $E_1$, and $E_2$, with $E_0=
E(\Tr(0))$,
\begin{itemize}
\item $E_1 := \{ e\in E(\Tr(n))\setminus E(\Tr(0))$: there exists some $i>0$,
$v\in V(\Tr(i))\setminus V(\Tr(i-1))$ such that $v$ is inside a triangle
$u_1u_2u_3u_1$ which bounds an inner face of $\Tr(i-1)$, $e=u_1v$, and
$\sigma(e)\ne \sigma(vu_2)$ and $\sigma(e)\ne \sigma(vu_3)\}$.
\item $E_2 := \{ e\in E(\Tr(n))\setminus E(\Tr(0))$: there exists some $i>0$,
$v\in V(\Tr(i))\setminus V(\Tr(i-1))$ such that $v$ is inside a triangle
$u_1u_2u_3u_1$ which bounds an inner face of $\Tr(i-1)$, $e=u_1v$ and
$\sigma(e)=\sigma(vu_2)$ or $\sigma(e)=\sigma(vu_3)\}$, and
\end{itemize}
\end{comment}
%$E_3$ := \{$e\in E(\Tr(n))\setminus$ $E(\Tr(0))$: $e=u_1v$, $v\in$ $V(\Tr(i))\setminus$
%$V(\Tr(i-1))$ for $i>0$ and $v$ is $inside$ a triangle $u_1u_2u_3u_1$, where
%$u_1u_2u_3u_1$ is a triangle bounding an inner face of $\Tr(i-1)$ s.t
%$\sigma(vu_1)$$\neq$$\sigma(vu_2)$ and $\sigma(vu_1)$$\neq$$\sigma(vu_3)$\}.
%We say an edge $e$ is of {\it type} $i$ if $e \in E_i$.
\begin{itemize}
\item [(2)] Let $v_0v_1v_2v_0$ be a triangle bounding an inner face of $\Tr(i)$, where $0\le i<n$,
let $v\in V(\Tr(i+1))\setminus V(\Tr(i))$
with $v$ inside $v_0v_1v_2v_0$. Then, for any $v_0w\in E(\Tr(n))$ with $w$ inside $v_0v_1v_2v_0$, $\sigma(v_0w)=\sigma(v_0v)$.
\end{itemize}
Let $v_0w\in E(\Tr(n))$ with $w$ inside $v_0v_1v_2v_0$. Then there exists $k\ge 0$ with $i+k+1\leq n$,
such that $w\in V(\Tr(i+k+1))\setminus V(\Tr(i+k))$. We prove (2) by
applying induction on $k$.
The basis case is trivial because $k=0$ implies $w=v$.
So assume $k \geq 1$. Let $v_0v_3v_4v_0$ be the triangle bounding an inner face of
$\Tr(i+k-1)$ with $w$ inside $v_0v_3v_4v_0$, and let $v_5 \in
V(\Tr(i+k))\setminus V(\Tr(i+k-1))$ that is inside $v_0v_3v_4v_0$. By symmetry, assume $w$ is inside $v_0v_5v_4v_0$.
By induction hypothesis, $\sigma(v_0v_5)$=$\sigma(v_0v)$.
Suppose $\sigma(v_4v_5)=\sigma(v_0v_5)$. %Then, by (1), $\sigma(v_0v_5)\ \neq\ \sigma(v_3v_5)$.
Hence by (ii) and (iii), $\sigma(v_0w)=\sigma(wv_4)=\sigma(v_0v_5)$.
Thus $\sigma(v_0w)=\sigma(v_0v)$.
Now assume $\sigma(v_4v_5)\neq\sigma(v_0v_5)$. Then $\sigma(v_3v_5) = \sigma(v_0v_5)$ or $\sigma(v_3v_5)$=$\sigma(v_4v_5)$.
It follows from (iii) that $\sigma(v_0w)=\sigma(v_0v_5)$. Hence, $\sigma(v_0w)=\sigma(v_0v)$. $\Box$
%(ii) and (iii), $\sigma(v_0w)=\sigma(v_0v_5)$; so $\sigma(v_0w)=\sigma(v_0v)$.
%If $\sigma(v_3v_5)$=$\sigma(v_4v_5)$ then, by
%(ii) and (iii), $\sigma(v_0w)$=$\sigma(v_5w)$=$\sigma(v_0v_5)$; so
%$\sigma(v_0w)$=$\sigma(v_0v)$. $\Box$
\begin{itemize}
\item [(3)] Let $v_0v_1v_2v_0$
be a triangle bounding an inner face of $\Tr(i)$ with
$0\le i\le n-2$, and let $v \in V(\Tr(i+1))\setminus V(\Tr(i))$ such that
$v$ is inside $v_0v_1v_2v_0$
and $\sigma(vv_0)\ne \sigma(vv_1)=\sigma(vv_2)$. Then for any $vw\in
E(\Tr(n))$ with $w$ inside $v_0v_1v_2v_0$,
$\sigma(vw)=\sigma(vv_0)$.
\end{itemize}
To prove (3), let $\{w_0,w_1,w_2\}\subseteq V(\Tr(i+2))\setminus V(\Tr(i+1))$ such that $w_j$ is inside $vv_jv_{j+1}v$
for $j=0,1,2$, with subscripts modulo 3. By (ii) and (iii), $\sigma(vw_0)=\sigma(vw_2)=\sigma(vw_1)=\sigma(vv_0)$.
By (2), there exists some $j\in \{0,1,2\}$ with $\sigma(vw)=\sigma(vw_j)$.
Hence, $\sigma(vw) = \sigma(vv_0)$. $\Box$
\begin{itemize}
\item [(4)] Let $v_0v_1v_2v_0$ be a triangle bounding an inner face of $\Tr(i)$, where $0\le i\le n-2$, and let
$v\in V(\Tr(i+1))\setminus V(\Tr(i))$ such that $v$ is inside $v_0v_1v_2v_0$ and $\sigma(vv_0)\in \{\sigma(vv_1),\sigma(vv_2)\}$.
Then for any $w\in (N(v)\cap N(v_0))\setminus \{v_1,v_2\}$, $\sigma(wv_0) \neq \sigma(wv)$.
%$vw\in E(\Tr(n))$ with $w\notin \{v_1,v_2\}$ and $wv_0,wv\in E(\Tr(n))$, $\sigma(wv_0) \neq \sigma(wv)$.
\end{itemize}
To prove (4), we may assume by symmetry and (1) that $\sigma(vv_2)\neq \sigma(vv_0)=\sigma(vv_1)$. Then $\sigma(wv_0)=\sigma(vv_0)$ by (2), and $\sigma(wv)=\sigma(vv_2)$ by (3). Hence, $\sigma(wv_0)\ne \sigma(wv)$. $\Box$
\begin{itemize}
\item [(5)] Suppose $upv$ is a monochromatic path of length two in $\Tr(n)$ with $uv\in E(\Tr(i+1))$ and $p\in V(\Tr(n))\setminus V(\Tr(i+1))$.
Then any monochromatic path in $\Tr(n)$ between $u$ and $v$ and of the color $\sigma(up)$ has length at most two.
%Suppose $uv\in E(\Tr(i+1))$ for some $i\geq 0$, $p \in V(\Tr(n))\setminus V(\Tr(i+1))$, $pu,pv\in E(\Tr(n))$, and $\sigma(pu)=\sigma(pv)$.
%Let $a_0a_1a_2\ldots a_r$ be a path in $\Tr(n)$ such that $a_0=v$, $a_r=u$, and
%$\sigma(a_ia_{i+1})=\sigma(pu)$ for $i=0, \ldots, r-1$. Then $r\leq 2$.
\end{itemize}
Consider any monochromatic path $P=a_0a_1...a_r$ of the color $\sigma(up)$ with $a_0=v$ and $a_r=u$.
First, suppose $uv\in E(\Tr(0))$. Let $\Tr(0)=uvwu$ and $x\in V(\Tr(1)) \setminus V(\Tr(0))$.
By (2), $\sigma(ux)=\sigma(up)$ and $\sigma(vx)=\sigma(vp)$; so $\sigma(xu)=\sigma(xv)$. Thus, by (i),
$\sigma(wx)=\sigma (wu)=\sigma(wv)\ne \sigma(xu)$.
Let $v_0v_1\ldots v_n$ be a path in $\Tr(n)$ with $v_0=w$, $v_1=x$ and for $1\le i\le n$,
$v_i\in V(\Tr(i))\setminus V(\Tr(i-1))$ is inside $v_{i-1}uvv_{i-1}$. By (ii) and (iii),
$\sigma(v_iu)=\sigma(v_iv)=\sigma(vx)$ for $1\leq i\leq n$, and
$\sigma(v_iv_{i+1})= \sigma(xw)$ for $0\le i\le n-1$.
By planarity, $P$ is contained in the closed region bounded by $uvwu$.
So either $P=uv$ or there exists some $1\leq k\leq r-1$ such that $a_k\in \{v_0,...,v_n\}$.
We may assume the latter case occurs.
If $\{a_{k-1}, a_{k+1}\}=\{u,v\}$, then $r=2$.
Hence without loss of generality, let $a_{k-1}\notin \{u,v\}$.
Then by (2) and (3), $\sigma(a_{k-1}a_k)=\sigma(v_iv_{i+1})\neq \sigma(pu)$ for $i\in \{0,1,...n-1\}$, a contradiction.
Hence $r\le 2$.\footnote{We remark that this paragraph also shows that such $uv$ in $E(\Tr(0))$ cannot be in a monochromatic $C_4$.}
\begin{comment}
Recall (i), only one of $ow,ou,ov$ are in $E_1$. If
$ow\in E_1$, we have $\sigma(ou)=\sigma(ov)=\sigma(uv)$,
$\sigma(ow)=\sigma(wu)=\sigma(wv)$.
Also, $\sigma(uw)=\sigma(vw) \neq \sigma(uv)$.
We then have a list of vertices
$v_1,\ldots, v_n$, where $v_1$ is $o$ and $v_i\in V(\Tr(i))\setminus V(\Tr(i+1))$ is
inside $v_{i-1}uvv_{i-1}$ ($1\leq i\leq n$). Let $v_0=w$. It is known that
$\sigma(v_iu)=\sigma(v_iv)=\sigma(uv)$ ($1\leq i\leq n$).
$\sigma(v_iv_{i+1})= \sigma(uw)= \sigma(vw)$($0\leq i\leq n-1$). Then we can prove
$r\leq 2$ the same as the proof in the second part of Case 1, a contradiction. So
by symmetry, we suppose $ou\in E_1$. Then $\sigma(ou) \neq \sigma(ov)$. By (2),
$\sigma(pu)=\sigma(ou)$, $\sigma(pv)=\sigma(ov)$. Hence,
$\sigma(pu) \neq \sigma(pv)$, a contradiction.
\end{comment}
Thus, we may assume $uv\notin E(\Tr(0))$. By symmetry, we may assume that
$v\in V(\Tr(i+1)) \setminus V(\Tr(i))$ for some $0\le i <n$ and $v$ is inside the triangle $u_1u_2u_3u_1$ bounding an inner
face of $\Tr(i)$ and $u_1=u$. By (4), $\sigma(u_1v)\neq \sigma(u_2v) =\sigma(u_3v)$.
If $a_1$ is inside $vu_2u_3v$ then there exists $1\le k<r$ such that $a_k$ is inside
$vu_3u_2v$ and $a_{k+1}\in \{u_2, u_3\}$; so by (2),
$\sigma(a_ka_{k+1})= \sigma(vu_2)= \sigma(vu_3) \neq \sigma(u_1v)=\sigma(pu)$, a contradiction.
Therefore, suppose that $P\neq uv$, by symmetry, we may assume that $a_1$ is inside $u_1vu_2u_1$.
Let $v_0=u_2$ and let $v_1v_2 \ldots v_{n-i-1}$ be the path in $\Tr(n)$ such that,
%$v_1 \in V(\Tr(i+2)) \setminus \Tr(i+1)$ is inside $u_1u_2vu_1$ and,
for $1\le \ell\le n-i-1$, $v_\ell\in V(\Tr(i+\ell+1)) \setminus V(\Tr(i+\ell))$ is inside $u_1v_{\ell-1}vu_1$.
By (ii) and (iii), $\sigma(v_\ell u_1)=\sigma(v_\ell v)=\sigma(u_1v)$ for
$1\leq \ell\leq n-i-1$, and $\sigma(v_\ell v_{\ell+1})=\sigma(vu_2) \neq \sigma(vu_1)$
for $0\leq \ell\leq n-i-2$.
%Note that either $a_1$ is in $v_\ellv_{l+1}vv_\ell$ for $0\leq l \leq h-1$ or $a_1\in \{v_1,v_2,\ldots, v_h\}$.
If $a_1$ is inside $v_\ell v_{\ell+1}vv_\ell$ for some $\ell$ with $0\le \ell\le n-i-2$, then exists $1\le k\le r$ such that
$a_k$ is inside $v_\ell v_{\ell+1}vv_\ell$ and $a_{k+1}\in \{v_\ell,v_{\ell+1}\}$;
so by (3) $\sigma(a_ka_{k+1}) = \sigma(v_\ell v_{\ell+1})$, a contradiction.
So $a_1=v_\ell$ for some $\ell$ with $1\le \ell\le n-i-1$.
Then as $\sigma(a_1a_2) = \sigma(u_1v)$ and by (3), we have $a_2=u_1$.
Therefore, $r=2$, proving (5). $\Box$
\medskip
\begin{itemize}
\item [(6)] If $C_k$ is monochromatic in $\Tr(n)$ then $k\leq 4$.
\end{itemize}
%We prove (6) by contradiction. Suppose there is a monochromatic $C_k$ in $\Tr(n)$ for
%$k\geq 5$, $V(C_k)=\{a_1,a_2,\ldots, a_k\}$, and $E(C_k)=\{a_ia_{i+1}\}$, where subscripts are
%taken modulo $k$.
Let $C_k=a_1a_2\ldots a_ka_1$ be a monochromatic cycle in $\Tr(n)$.
By (i), $E(C_k)\not \subseteq E(\Tr(0))$.
So we may assume that there exists some $1\le i\le k$ such that $a_{i+1} \in V(\Tr(\ell+1))\setminus V(\Tr(\ell))$ is inside the triangle $a_iuva_i$
which bounds an inner face of some $\Tr(\ell)$.
We may further assume that $\ell\le n-2$, as otherwise, we could consider
$\Tr(n+1)$ instead of $\Tr(n)$.\footnote{This is fair because $\Tr(n+1)\not\to C_k$ implies $\Tr(n)\not\to C_k$.}
Suppose $\sigma(a_ia_{i+1})\in \{\sigma(a_{i+1}u),\sigma(a_{i+1}v)\}$. By symmetry, we may assume
$\sigma(a_ia_{i+1})= \sigma(a_{i+1}u)$. Then $a_{i+2}=u$ by (3).
Hence, by (5), any monochromatic path in $C_k$ between $a_i$ and $a_{i+2}=u$ has length at most 2.
So $k\leq 4$.
Thus, we may assume $\sigma(a_ia_{i+1})\notin \{\sigma(a_{i+1}u),\sigma(a_{i+1}v)\}$; hence, $\sigma(a_{i+1}u)=\sigma(a_{i+1}v)$.
Let $w\in V(\Tr(\ell+2)) \setminus V(\Tr(\ell+1))$ be inside the triangle
$a_iua_{i+1}a_i$. By (ii) and (iii), $\sigma(wa_i) = \sigma(wa_{i+1}) = \sigma(a_ia_{i+1})$.
Hence, by (5), the monochromatic path $C_k-a_ia_{i+1}$ in $\Tr(n)$ of the color $\sigma(a_ia_{i+1})=\sigma(wa_i)$ has length at most 2; so $k=3$. $\Box$
\begin{comment}
%First, suppose $E(C_{k})\subseteq E_0\cup E_2$.
%First, suppose for some $i$, $a_ia_{i+1}\in E_2$. In addition, $a_iuva_i$ is a triangle bounding an
%inner face of $\Tr(\ell)$ and $a_{i+1} \in V(\Tr(\ell+1)) \setminus V(\Tr(\ell))$ is in
%$a_iuva_i$. By symmetry, suppose $a_{i+1}u\in E_2$ so $a_{i+1}v\in E_1$. By (3)
%$a_{i+2}=u$. Then by (5) any monochromatic path $v_0=a_i,v_2,\ldots, v_\ell=u$, where
%$\sigma(v_hv_{h+1})=\sigma(a_ia_{i+1})$ ($l-1 \geq h \geq 0$) satisfies $l\leq 2$.
%Hence $k \leq 4$, a contradiction.
So $E(C_{k})\cap E_1\ne \emptyset$. Let $a_ia_{i+1}\in E_1$. In addition,
$a_iuva_i$ is a triangle bounding an inner face of $\Tr(\ell)$ and
$a_{i+1} \in V(\Tr(\ell+1)) \setminus V(\Tr(\ell))$ is inside $a_iuva_i$.
Then $a_{i+1}u, a_{i+1}v\in E_2$. We assume $l+1\leq n-1$ because the coloring method is
iterative and $Tr(s)$ is a subgraph of $Tr(s+1)$, where the color of the
corresponding edge is the same. $w\in V(\Tr(\ell+2)) \setminus V(\Tr(\ell+1))$ is inside
$a_1ua_2a_1$. By (ii) and (iii), $\sigma(wa_1) = \sigma(wa_2) = \sigma(a_1a_2)$.
Similarly, by (5), $k\leq 3$, a contradiction. $\Box$
\end{comment}
\begin{itemize}
\item [(7)] There is no monochromatic $H^+$ in $\Tr(n)$.
\end{itemize}
Suppose that there is a monochromatic copy of $H^+$ on $\{v_i: 1\le i\le 6\}$ in which
$v_1v_2v_3v_4v_1$ is a 4-cycle and $v_1v_5,v_2v_6$ are edges.
If $v_1v_2\in E(\Tr(0))$, then $v_1v_2$ satisfies the conditions of (5) and by the footnote from the proof of (5),
there is no monochromatic $C_4$ containing $v_1v_2$, a contradiction.
So $v_1v_2\notin E(\Tr(0))$. By symmetry, we may assume that
$v_2\in V(\Tr(i+1))\setminus V(\Tr(i))$ for some $i$ and that $v_1uwv_1$ is the triangle bounding the
inner face of $\Tr(i)$ containing $v_2$.
Again as before we may assume that $0\leq i\leq n-2$.
If $\sigma(v_2u)=\sigma(v_2w)$, then there exists some $p\in V(\Tr(n))\setminus V(\Tr(i+1))$ such that
$v_1pv_2$ has the same color as $\sigma(v_1v_2)$.
But $v_1v_4v_3v_2$ is a monochromatic path of length 3 in $\Tr(n)$ between $v_1$ and $v_2$ and of the color $\sigma(v_1v_2)$, a contradiction to (5).
Hence, $\sigma(v_1v_2)\in \{\sigma(v_2u), \sigma(v_2w)\}$ and by symmetry, we may assume $\sigma(v_1v_2)=\sigma(v_2u)$.
Then by (1), $\sigma(v_1v_2)\neq\sigma(v_2w)$ and thus $\sigma(v_2v_3)=\sigma(v_2v_6)\ne \sigma(v_2w)$.
This shows $w\notin \{v_3,v_6\}$. So there exists $y\in \{v_3,v_6\}\setminus \{u,w\}$.
By (3), $\sigma(v_2y)=\sigma(v_2w)$, a contradiction.
This completes the proof of this section.
\begin{comment}
Suppose G is monochromatic. Then by the proof of (5). $\{v_1v_2,\ v_2v_3,\ v_3v_5,\
v_5v_1\}\subset E_2$. Otherwise if $e\in \{v_1v_2,\ v_2v_3,\ v_3v_5,\ v_5v_1\}$ is
in $E_1$ or $E_0$, the length of a monochromatic $C_k$, which contains $e$ as an
edge satisfies $k\leq 3$. Contradicts. Therefore, $v_1v_2\in E_2$.
With no loss of
generality, suppose $v_1uwv_1$ bounds an inner face of $\Tr(i)$($i<n$) and $v_2\in
V(\Tr(i+1))\setminus V(\Tr(i))$ is inside $v_1uwv_1$. Because $v_1v_2\in E_2$, by
symmertry, suppose $\sigma(v_1v_2)=\sigma(v_2u)$. Then
$\sigma(v_1v_2)\neq\sigma(v_2w)$. By (3), $\forall x$ inside $v_1uwv_1$ and
connected to $v_2$ satisfies $\sigma(xv_2)=\sigma(v_2w)$. In that, at least one
vertex y in $\{v_3,v_6\}$ satisfies $\sigma(yv_2)=\sigma(v_2w)\neq \sigma(v_1v_2)$, a contradiction. $\Box$
\end{comment}
\section{Monochromatic $K_{2,3}$}
In this section, we prove Theorem~\ref{main} for $K_{2,3}$ using a different coloring scheme on $\Tr(n)$ described below. Let $\sigma: E(\Tr(n))\to \{0,1\}$ be
defined inductively as follows:%For each $e\in E(\Tr(n))$, let
\begin{itemize}
\item [(i)] Fix a triangle $T$ bounding an inner face of $\Tr(1)$, and let $\sigma(e)=0$ if $e\in E(T)$ and $\sigma(e)=1$
if $e\in E(\Tr(1))\setminus E(T)$. %(See Figure 3.)
\item [(ii)] Suppose for some $1\le i<n$, we have defined $\sigma(e)$ for all $e\in E(\Tr(i))$. We now extend $\sigma$ to $E(\Tr(i+1))$.
Let $x\in V(\Tr(i))\setminus V(\Tr(i-1))$ be arbitrary, let $v_0v_1v_2v_0$ denote the triangle bounding
the inner face of $\Tr(i-1)$ containing $x$, with $v_0,v_1,v_2$ on the triangle in clockwise order, and let
$\sigma(xv_1)=\sigma(xv_2)$.
% Moreover, suppose $v_0,v_1,v_2$ appear clockwise.
%We ensure this order just to ensure the coloring to be unique, however the
%proof has no connection to it.
\item [(iii)] Let $x_j\in V(\Tr(i+1))\setminus V(\Tr(i))$ such that
$x_j$ is inside the face of $\Tr(i)$ bounded by the triangle $xv_jv_{j+1}x$,
where $j=0,1,2$ and the subscripts are taken modulo 3. Define $\sigma(v_0x)=\sigma(v_0x_0)=\sigma(v_0x_2)
=\sigma(xx_2) =\sigma(x_1v_1)$, and $\sigma(v_2x)=\sigma(v_2x_1)=\sigma(v_2x_2)=\sigma(xx_1)
=\sigma(xx_0)=\sigma(x_0v_1)$.% (See Figure 5)
\end{itemize}
%\begin{figure}
% \centering
% \includegraphics[scale=0.5]{16.png}
% \caption{new iterative coloring}
%\end{figure}
Note that in (ii) we have $|\{\sigma(xv_j): j=0,1,2\}|=2$ and that in (iii) we have $\sigma(x_jv_j)\ne \sigma(x_jv_{j+1})$ for $j=0,1,2$.
Hence, inductively, we have
\begin{itemize}
\item [(1)] For $1\le i\le n$ and $x\in V(\Tr(i))\setminus V(\Tr(i-1))$,
$|\{\sigma(xv): v\in V(\Tr(i-1))\}|=2$.
\item [(2)] If $x_1x_2x_3x_1$ is a triangle which bounds an inner face of $\Tr(i)$ for some $1\le i\le n-2$, and if $x\in
V(\Tr(n))\setminus V(\Tr(i+1))$ is inside $x_1x_2x_3x_1$ with $xx_1,xx_2\in E(\Tr(n))$, then $\sigma(xx_1)\neq \sigma(xx_2)$.
\end{itemize}
These two claims are straightforward so we omit their proofs.
\begin{itemize}
\item [(3)] For any $x_1x_2\in E(\Tr(n))$, $|\{x\in
N(x_1)\cap N(x_2): \sigma(xx_1)=\sigma(xx_2)=0\}|\le 2$ and $|\{x\in N(x_1)\cap
N(x_2): \sigma(xx_1)=\sigma(xx_2)=1\}|\le 2$.
\end{itemize}
First, suppose $x_1x_2\in E(\Tr(0))$. Then by (i) and (2),
$|\{x\in N(x_1)\cap N(x_2): \sigma(xx_1)=\sigma(xx_2)=0\}|\le 1$ and $|\{x\in N(x_1)\cap
N(x_2): \sigma(xx_1)=\sigma(xx_2)=1\}|\le 1$.
So we may assume that $x_1vwx_1$ bounds an inner face of $\Tr(i)$ and $x_2\in V(\Tr(i+1))\setminus V(\Tr(i))$ inside $x_1vwx_1$.
%Moreover, assume $x_1,v,w$ occur on the triangle in clockwise order.
Let $v_1\in \Tr(i+2)$ be inside $x_1vx_2x_1$ and $w_1\in \Tr(i+2)$ be inside $x_1wx_2x_1$. By (iii), $\sigma(w_1x_1)\neq \sigma(w_1x_2)$ or $\sigma(v_1x_1)\ne \sigma(v_1x_2)$.
By (2), for any $x\in V(\Tr(n))\setminus V(\Tr(i+2))$ inside $x_1vwx_1$ with $xx_1,xx_2\in E(\Tr(n))$, we have $\sigma(xx_1)\neq \sigma(xx_2)$.
Hence, if (3) fails, then
we may assume by symmetry between $w_1$ and $v_1$ that $\sigma(vx_1)=\sigma(vx_2)=\sigma(wx_1)=\sigma(wx_2)=\sigma(v_1x_1)=\sigma(v_1x_2)$, and $\sigma(w_1x_1)\neq \sigma(w_1x_2)$.
Then, by (1), $\sigma(x_1x_2)\neq \sigma(x_2v)=\sigma(x_2w)$.
Now by (iii), at least one of the two edges $v_1x_1$ and $v_1x_2$ has the same color as $x_1x_2$, a contradiction. This proves (3).
%(iii), $\sigma(v_1x_1)=\sigma(v_1x_2)=\sigma(x_1x_2)$ and, hence, $\sigma(x_1x_2) =\sigma(v_1x_1)=\sigma(vx_2)$, a contradiction. $\Box$
\begin{itemize}
\item [(4)] If $x_1x_2x_3x_4x_1$ is a 4-cycle in $\Tr(n)$, then $x_1x_3\in E(\Tr(n))$ or $x_2x_4\in E(\Tr(n))$.
\end{itemize}
We may assume that $\{x_1,x_2,x_3,x_4\}\subseteq V(\Tr(i+1))$ and $x_j\in V(\Tr(i+1))\setminus V(\Tr(i))$ for some $0\le i<n$ and $j\in [4]$.
Let $uvwu$ be the triangle bounding an inner face of $\Tr(i)$ such that $x_j$ is inside it.
Then $\{x_{j-1},x_{j+1}\}\subseteq \{u,v,w\}$, implying that $x_{j-1}x_{j+1}\in E(\Tr(n))$. $\Box$.
\begin{itemize}
\item [(5)] There is no monochromatic $K_{2,3}$ in $\Tr(n)$.
\end{itemize}
For, suppose $\Tr(n)$ has a monochromatic copy of $K_{2,3}$ on $\{v_1,v_2,v_3,v_4,v_5\}$ with $v_4v_i, v_5v_i\in E(\Tr(n))$ for all $i=1,2,3$.
%$G[\{v_1,v_2,v_3,v_4\}]$=$v_1v_2v_3v_4v_1$ is a $C_4$. $N_G(v_5)=\{v_2,v_4\}$. By
Then $v_4v_5\notin E(\Tr(n))$ by (3) and, hence, it follows from (4) that $v_1v_2,v_2v_3,v_3v_1\in E(\Tr(n))$. By planarity,
$v_1v_2v_3v_1$ bounds an inner face of $\Tr(i)$ for some $i$ with $1\le i<n$ and, by the symmetry between $v_4$ and $v_5$, we may
assume that $v_4$ is inside $v_1v_2v_3v_1$. Then $v_4\in V(\Tr(i+1))\setminus V(\Tr(i))$.
However, this contradicts (1), as $\sigma(v_4v_1)=\sigma(v_4v_2)=\sigma(v_4v_3)$.
We have completed the proof of Theorem \ref{main}. \qed
\section{Monochromatic $J_k$}
In this section we prove that $\Tr(100k)\to J_k$ holds for any positive integer $k$.
We need the following result, which is Lemma 9 in \cite{ASTU19}.
The original statement in \cite{ASTU19} states $\Tr(16)\to C_4$,
but the same proof in \cite{ASTU19} actually gives the following stronger version.
\begin{lem}\label{astu}
If $xyzx$ bounds the outer face of $\Tr(16)$, then any 2-edge-coloring of $\Tr(16)$ gives a monochromatic $C_4$
which intersects $\{x,y\}$.
\end{lem}
Note that if the triangle $xyzx$ bounds the outer face of $\Tr(n)$ and $v\in V(\Tr(1))\setminus V(\Tr(0))$ then the subgraph of $\Tr(n)$ contained in the
closed disc bounded by $vxyv$ is isomorphic to $\Tr(n-1)$. Hence, the following is an easy consequence of Lemma~\ref{astu}.
\begin{coro}\label{l1}
If $xyzx$ bounds the outer face of $\Tr(17)$ then any 2-edge-coloring of $\Tr(17)$ gives a monochromatic $C_4$ which intersects $\{x,y\}$ and avoids $z$.
\end{coro}
%\pf By Lemma~\ref{astu},
%let $x_1\in V(\Tr(1))\setminus V(\Tr(0))$. Then
%$xx_1yx$ bounds the outer face of a $Tr(16)$. The conclusion holds by applying Lemma~\ref{astu}.\qed
\begin{lem}\label{l2}
For any positive integer $k$, $\Tr(38k)$$\rightarrow F_k$
\end{lem}
\begin{proof}
Let $\sigma: E(\Tr(38k))\to \{0,1\}$ be an arbitrary 2-edge coloring.
Let $uvwu$ be the triangle bounding the outer face of $\Tr(38k)$. Let $x_0:=w$ and, for $1\le \ell\le 2k$, let
$x_l\in V(\Tr(\ell))\setminus V(\Tr(\ell-1))$ such that $x_\ell$ is inside $x_{\ell-1}uvx_{\ell-1}$.
Let $y_{i,0}:=x_i$ for $i\in \{0,1, \ldots, 2k-1\}$ and, for $\ell\in \{1, \ldots , 36k\}$,
let $y_{i,\ell}\in V(\Tr(i+1+\ell))\setminus V(\Tr(i+\ell))$ such that $y_{i,\ell}$ is inside $y_{i,\ell-1}ux_{i+1}y_{i,\ell-1}$.
%(See Figure 6.)
%To prove the assertion, we consider any 2-edge-coloring $\sigma: E(\Tr(36k))\to \{0,1\}$.
%Let $x_i$ be $y_{i,0}$.\\
Suppose for each $0\leq i\leq 2k-1$ there exists a monochromatic $C_4$ inside $x_iux_{i+1}x_i$ that contains $u$
and avoids $x_i$. By pigeonhole principle, at least $k$ of these $C_4$s are of the same color, which form a
monochromatic $F_k$ centered at $u$.
Hence, we may assume that there exists some $i\in \{0,1,\ldots, 2k-1\}$ such that no monochromatic
$C_4$ inside $x_iux_{i+1}x_i$ contains $u$ and avoids $x_i$.
Since $i\le 2k-1$, $x_iux_{i+1}x_i$ bounds the outer face of a $\Tr(36k)$ that is contained in $\Tr(38k)$.
Now for each $h\in \{0, 1, \ldots, 2k-1\}$, we view the closed region bounded by $ux_{i+1}y_{i,18h}u$ as a $\Tr(17)$.
Note that these copies of $\Tr(17)$ share $u, x_{i+1}$ as the only common vertices.
Taking $y_{i,18h}$ to be the vertex $z$ in Corollary \ref{l1}, we conclude from Corollary \ref{l1} that
there is a monochromatic $C_4$ in the $\Tr(17)$ bounded by $ux_{i+1}y_{i,18h}u$ such that $x_{i+1}\in V(G_h)$ and $\{u,y_{i,18h}\}\cap V(G_h)=\emptyset$.
By pigeonhole principle, at least $k$ of these $C_4$s are of the same color,
which clearly form a monochromatic $F_k$ centered at $x_{i+1}$.
\end{proof}
%\begin{figure}
% \centering
% \includegraphics[scale=0.5]{13.png}
% \caption{lemma \ref{l2}}
%\end{figure}
\begin{lem}\label{l3}
Let $k$ be a positive integer and let $uvwu$ bound the outer face of $\Tr(9k+2)$.
Suppose $\sigma: E(\Tr(9k+2))\to \{0,1\}$ is a 2-edge-coloring such that $|\{\sigma(ux): x\in V(\Tr(9k+2))\}|=1$ and
there is no monochromatic $C_4$ containing $u$. Then
$\Tr(9k+2)$ contains monochromatic $J_k$ centered at $v$.
\end{lem}
\begin{proof} Without loss of generality, assume $\sigma(uv)=0$. Then $\sigma(uy)=0$ for all $y\in N(u)$.
Let $x_0:=w$ and, for $1\le i\le 8k+1$, let $x_i\in V(\Tr(i))\setminus V(\Tr(i-1))$ such that $x_i$ is inside $x_{i-1}uvx_{i-1}$.
Since no monochromatic $C_4$ in $\Tr(9k+2)$ contains $u$, there is at most one $i\in \{0,1,2,...8k+1\}$ such that
$\sigma(x_iv)=0$. Hence, there exists $i\in \{0,1,\ldots 4k+2\}$ such that $\sigma(vx_j)=1$ for $j\in \{i,i+1\ldots i+4k-1 \}$.
We now make the following claim.
%\begin{itemize}
% \item [(1)] there exists $P_{4k}$ such that $V(P_{4k})\subseteq N(u)\cap N(v)$ and, for $x\in V(P_{4k})$, $\sigma(xu)=0$ and $\sigma(xv)=1$.
%\end{itemize}
% Since no monochromatic $C_4$ in $\Tr(9k+2)$ contains $u$, there is at most one $i\in \{0,1,2,...8k+1\}$ such that
%$\sigma(x_iv)=0$. Therefore, there exists $i\in \{0,1,\ldots 4k+2\}$ such that $\sigma(vx_j)=1$ for $j\in \{i,i+1\ldots i+4k-1 \}$.
\begin{itemize}[leftmargin=14mm]
\item[Claim.] The subgraph of $\Tr(9k+2)$ contained in the closed disc bounded by $vx_i\ldots x_{i+3k-1}v$ has
a monochromatic $F_k$ of color 1 and centered at $v$, which we denote by $F_v$.
\end{itemize}
To show this, it suffices to show that for each $r$ with $0\le r\le k-1$,
the subgraph $T_r$ of $\Tr(9k+2))$ inside $vx_{i+3r}x_{i+3r+1}x_{i+3r+2}v$ (inclusive)
contains a monochromatic $C_4$ of color 1 and containing $v$, as the union of such $C_4$ is an $F_k$ centered at $v$.
So fix an arbitrary $r$, with $0\le r\le k-1$.
Note that $\sigma(x_{i+3r}x_{i+3r+1})=1$ or
$\sigma(x_{i+3r+1}x_{i+3r+2})=1$, for $0\leq r\leq k-1$; for, otherwise, $x_{i+3r}x_{i+3r+1}x_{i+3r+2}ux_{i+3r}$ is a monochromatic $C_4$
of color 0 and containing $u$, a contradiction. Without loss of generality, assume
$\sigma(x_{i+3r}x_{i+3r+1})=1$.
Let $y\in V(\Tr(i+3r+2))\setminus V(\Tr(i+3r+1))$ such that $y$ is inside
$x_{i+3r}x_{i+3r+1}vx_{i+3r}$.
If there are two edges in $\{yx_{i+3r},yx_{i+3r+1},yv\}$ of color 0,
then one can easily find a monochromatic $C_4$ of color $0$ and containing $u$, a contradiction.
%If $\sigma(yx_{i+3r})=\sigma(yx_{i+3r+1})=0$ then
%$ux_{i+3r}yx_{i+3r+1}u$ is monochromatic $C_4$ containing $u$, a contradiction.
%If $\sigma(yx_{i+3r})=\sigma(yv)=0$ then $x_{i+3r}yvux_{i+3r}$ is
%a monochromatic $C_4$ containing $u$, a contradiction. If
%$\sigma(yx_{i+3r+1})=\sigma(yv)=0$ then $yvux_{i+3r+1}y$ is a
%monochromatic $C_4$ containing $u$, also a contradiction.
Hence, at least two of $\{\sigma(yx_{i+3r}),\sigma(yx_{i+3r+1}),\sigma(yv)\}$ are 1.
So $\{y,x_{i+3r},x_{i+3r+1},v\}$ induces a subgraph which contains a monochromatic $C_4$ of color 1.
This proves the claim. $\Box$
\begin{comment}
Hence, we may assume $\sigma(x_{i+3r+1}x_{i+3r+2})=1$. Let
$y\in V(\Tr(i+3r+3))\setminus V(\Tr(i+3r+2))$ such that $y$ is inside
$x_{i+3r+1}x_{i+3r+2}vx_{i+3r+1}$.
By the same argument above, $\{x_{i+3r+1},x_{i+3r+2},y,v\}$ induces a subgraph which contains a
monochromatic $C_4$ of color 1.
So for $0\le r\le k-1$, let $G_r$ denote a monochromatic $C_4$ of color 1 with vertex set either $\{y,x_{i+3r},x_{i+3r+1},v\}$ or $\{x_{i+3r+1},x_{i+3r+2},y,v\}$.
Note that $\cap^{k-1}_{r=0}V(G_r)=\{v\}$ and $|\cup^{k-1}_{r=0}V(G_r)|=3k+1$.
Clealry, $G_0, G_1, \ldots, G_{k-1}$ form a monochromatic
$F_k$ of color 1 centered at $v$.
\end{comment}
\medskip
%\begin{itemize}
% \item [(3)] There exists a monochromatic $J_k$ of color 1 centered at v.
%\end{itemize}
%Let $F_v$ denote the monochromatic $F_k$ centered at $v$. We now find a monochromatic perfect $k$-ary tree of radius 2 rooted at $v$.
%\begin{itemize}
% \item [(3)] Find a monochromatic $k-ary-perfect$ tree with radius 2 and root
% $v$, which together with $F_v$ form a monochromatic $J_k$ centered at $v$.
%\end{itemize}
%Consider $x_r\in \{x_{i+3k},x_{i+3k+1},...,x_{i+4k-1}\}$.
%We now find a monochromatic perfect $k$-ary tree of radius 2 rooted at $v$ but otherwise disjoint from $F_v$.
Note that for $i+3k\le r\le i+4k-1$, $ux_rx_{r-1}u$ bounds
the outer face of a $\Tr(k+1)$. Let $z_{r,0}:=x_{r-1}$ and, for $r \in \{i+3k,i+3k+1,\ldots i+4k-1\}$
and $\ell\in \{1,2,\ldots k\}$, let $z_{r,\ell}\in V(Tr(r+\ell))\setminus V(Tr(r+\ell-1))$
such that $z_{r,\ell}$ is inside $z_{r,\ell-1}x_ruz_{r,\ell-1}$.
Because $\sigma(uz_{r,j})=0$ (by assumption) and $\Tr(9k+2)$ has no monochromatic $C_4$ containing $u$, there is
at most one $y\in \{z_{r,1},z_{r,2},\ldots,z_{r,k}\}$ such that $\sigma(yx_r)=0$. So
there exists $k-1$ vertices in $\{z_{r,1},..., z_{r,k}\}$ which together with
$x_rv$ form a monochromatic $K_{1,k}$ of color 1 centered at $x_r$, which we denote by
$H_r$.
Now $H_{i+3k}, H_{i+3k+1}, \ldots, H_{i+4k-1}$ form a monochromatic $k$-ary radius 2 tree rooted at $v$ of color 1.
This radius 2 tree and $F_v$ form a monochromatic $J_k$ of color 1, completing the proof of Lemma \ref{l3}.
\end{proof}
\begin{comment}
\begin{lem}\label{l4}
Let $k,n$ be positive integers such that $n>2k$ and suppose $uvwu$ bounds the outer face of $\Tr(n)$.
For $i\in \{0,1\}$, let $A_i=\{x\in V(\Tr(n)): xu\in E(\Tr(n)) \mbox{ and } \sigma(xu)=i\}$. Then for $i\in \{0,1\}$,
$|A_i|<k$ implies that there exist vertices $w_1, w_2$ inside $uvwu$ such that
$uw_1w_2u$ bounds an inner face of $Tr(j)$ ($1\leq j\leq 2k$) and $\sigma(xu)=(1-i)$ when $x$ is inside $uw_1w_2u$ or $x\in \{w_1,w_2\}$.
\end{lem}
\begin{comment}
Suppose $uvwu$ bounds an inner face of $\Tr(i)$\in \Tr(n)(n>i+2k)$. $A_0$:=\{$x\in
V(\Tr(n))$: $x$ is inside $uvwu$. $x$ is connected to $u$ and
$\sigma(xu)=0$\}.
$A_1$:=\{$x\in V(\Tr(n))$: $x$ is inside $uvwu$. $x$ is connected to $u$ and
$\sigma(xu)=1$\}. If $|A_0|<k$, there will exist $w_1,\ w_2$
inside $uvwu$ s,t $uw_1w_2u$ bounds an inner face of $Tr(j)$($i+1\leq
j\leq i+2k$) and $\forall x$ inside $uw_1w_2u$ or $\in \{w_1,w_2\}$,
$\sigma(xu)=1.$ Similarly, if $|A_1|<k$, there will exist $w_1,\ w_2$
inside $uvwu$ s,t $uw_1w_2u$ bounds an inner face of $Tr(j)$($i+1\leq
j\leq i+2k$) and $\forall x$ inside $uw_1w_2u$ or $\in \{w_1,w_2\}$,
$\sigma(xu)=0.$
\pf Let $x_0:=w$. For $r\in [2k]$, let $x_r\in V(Tr(r))\setminus V(Tr(r-1))$ such that $x_r$ is
inside $uvx_{r-1}u$. By symmertry, suppose $|A_0|<k$. Then there exists $r\in [k]$ such that $\sigma(yu)=1$
for all $y\in N(u)$ such that $y$ is inside $ux_{2r-1}x_{2r}u$ or $y\in \{x_{2r-1},x_{2r}\}$. Therefore, $x_{2r-1},x_{2r}$ give the
required $w_1,w_2$. \qed
\end{comment}
Now we are ready to prove the main result of this section, that is $\Tr(100k)\to J_k$.
Let $\sigma: E(\Tr(100k))\to \{0,1\}$ be arbitrary. We show that $\sigma$ always contains a monochromatic $J_k$.
%\begin{itemize}
% \item [(1)] Find a monochromatic $F_{2k}$, which may be extended to a monochromatic $J_k$
%\end{itemize}
By Lemma \ref{l2}, $\Tr(76k)$ contains monochromatic copy of $F_{2k}$, say $F$, and, without loss of generality,
assume it is of color 1.
Let the $C_4$s in $F$ be $xa_{i,1}a_{i,2}a_{i,3}x$ for $i \in [2k]$.
For $i\in [2k]$, let $b_i\in V(\Tr(76k+1))\setminus V(\Tr(76k))$ such that $b_i$ is inside $xa_{i,1}a_{i,2}a_{i,3}x$ and
$a_{i,1}a_{i,2}b_ia_{i,1}$ bounds an inner face of $\Tr(76k+1)$.
Let $A_i$ be the family of all vertices $a\in N(a_{i,1})$ inside $a_{i,1}a_{i,2}b_ia_{i,1}$ and satisfying $\sigma(aa_{i,1})=1$.
\begin{itemize}
\item [(1)] There exists some $i\in \{k+1,k+2,...2k\}$ such that $|A_i|<k$.
%If the $F_{2k}$ can not be extended to a monochromatic $J_k$, we
% can use lemma\ref{l4} to find a subgraph $\in \Tr(100k)$, which satisfies the
% prerequisities of lemma\ref{l3}.
\end{itemize}
Otherwise, suppose $|A_i|\geq k$ for $i\in \{k+1,k+2,...2k\}$. Then let
$Z_{i}:=\{z_{i,1},z_{i,2},...z_{i,k-1}\}\subseteq A_i$. Now, for each $i\in \{k+1, \ldots, 2k\}$,
$\{x,a_{i,1}\}\cup Z_i$ induces a graph containing a monochromatic $K_{1,k}$.
Those $K_{1,k}$s form a monochromatic radius-two $k$-ary tree of color 1 and rooted at $x$, which we denote by $T_x$.
%Now $x$ and $xa_{i,1}a_{i,2}a_{i,3}x$ for $1\leq i\leq k$ form a monochromatic $F_k$ with color 1.
Now $F_k\cup T_x$ is a monochromatic $J_k$. $\Box$
\medskip
%Case2: There exists $i\in \{k+1,k+2,...2k\}$ s.t $|A_i|<k$. Then by
%Let $a_{i,1}$ and $u_0=a_{i,2}$, and, for $j\in [2k]$,
%let $u_j\in V(Tr(j))\setminus V(Tr(j-1))$ such that $u_j$ is inside $a_{i,1}b_iu_{j-1}a_{i,1}$.
Let $u:=a_{i,1}$. By (1), there exists an edge $vw\in \Tr(78k)$ such that $uvwu$ bounds an inner face of $\Tr(78k)$
and $\sigma(uy)=0$ for any $y\in N(u)$ in the closed disc bounded by $uvwu$.
Let $G$ be the subgraph of $\Tr(n)$ contained in the closed disc bounded by $uvwu$.
Clearly $G$ is isomorphic to a copy of $\Tr(22k)$.
In the rest of the proof, we should only discuss the graph $G$ and all $\Tr(i)$ will be referred to this copy of $\Tr(22k)$.
Let $x_0:=w$ and for $i\in [4k]$, let $x_{i}\in V(\Tr(i))\setminus V(\Tr(i-1))$ such that $x_{i}$ is inside $ux_{i-1}vu$.
\begin{itemize}
\item [(2)] $G$ contains a monochromatic copy of $F_k$, say $F'$, which has color 0 and center $u$ and is disjoint from the union of closed regions bounded by $ux_ix_{i+1}u$ over all $0\leq i\leq 2k-1$.
\end{itemize}
If for each $i\in \{k,k+1,...,2k-1\}$ there exists a monochromatic $C_4$ inside $ux_{2i}x_{2i+1}u$ and containing $u$,
then these $k$ monochromatic $C_4$s of color 0 form a desired monochromatic $F_k$ centered at $u$ and thus (2) holds.
Otherwise, since $ux_{2i}x_{2i+1}u$ bounds the outer face of a $\Tr(9k+2)$,
it follows from Lemma~\ref{l3} that there exists a monochromatic $J_k$ in $G$. $\Box$
\medskip
For $j\in \{0,1,...,2k-1\}$, let
$B_j$ be the family of all vertices $x\in N(x_j)$ inside $ux_jx_{j+1}u$ and satisfying $\sigma(xx_j)=0.$
\begin{itemize}
\item [(3)] There exists some $j\in \{0,1,..., k-1\}$ such that $|B_j|<k$.
\end{itemize}
Suppose on the contrary that there exist subsets $Z_j\subseteq B_j$ of size $k$ for all $j\in \{0,1,...,k-1\}$.
Then each $Z_j\cup \{u, x_j\}$ induces a graph containing a monochromatic $K_{1,k}$ which is centered at $x_j$ and has color 0.
These $K_{1,k}$s together with $F'$ form a monochromatic $J_k$ of color 0.
This proves (3). $\Box$
\medskip
Let $p_0:=x_{j+1}$ and for $1\leq \ell\leq 4k$, let
$p_\ell\in V(\Tr(j+\ell+1))\setminus V(\Tr(j+\ell))$ such that $p_\ell$ is inside $ux_jp_{\ell-1}u$.
By (3), there exists some $0\leq \ell\leq 4k-1$ such that $\sigma(zx_j)=1$ for any $z\in N(x_j)$ in the closed disc bounded by $x_jp_\ell p_{\ell+1}x_j$.
\begin{itemize}
\item [(4)]
%If $F_k^1$ cannot be extended to a monochromatic $J_k$ easily,
There is a monochromatic $F_k$ inside $x_jp_\ell p_{\ell+1}x_j$, say $F''$, with color 1 and center $x_j$.
\end{itemize}
Let $z_0:=p_{\ell+1}$ and for $s\in [2k]$, let $z_s\in V(\Tr(j+\ell+s+2))\setminus V(\Tr(j+\ell+s+1))$
such that $z_s$ is inside $x_jz_{s-1}p_\ell x_j$.
Note that each $x_jz_{2s}z_{2s+1}x_j$ bounds a $\Tr(9k+3)$.
If for each $s\in [k]$ there exists a monochromatic $C_4$ of color 1 inside $x_jz_{2s}z_{2s+1}x_j$ and containing $x_j$,
then these monochromatic copies of $C_4$ form the desired monochromatic $F_k$ centered at $x_j$.
Otherwise, it follows from Lemma \ref{l3} that there exists a monochromatic $J_k$. $\Box$
\medskip
As $|B_j|<k$, there exists a subset $A\subseteq \{p_1,p_2,...,p_{4k}\}$ of size $2k$ such that $\sigma(\alpha x_j)=1$ for each $\alpha\in A$
and moreover, there is no neighbors of $A$ belonging to $V(F'')$.
Let $A:=\{\alpha_1,..., \alpha _{2k}\}$.
Note that for each $h\in [2k]$, we have $\sigma(\alpha_h u)=0$ and $\sigma(\alpha_h x_j)=1$.
It is easy to see that there exist pairwise disjoint sets $N_h\subseteq N(\alpha_h)$ of size $2k$ for $h\in [2k]$.
%are connected to $p'_h$(It is obvious that $100k$ are large enough to guarantee the existence of $B_h$ for $h\in [2k]$).
Then there exists $M_h\subseteq N_h$ such that $|M_h|= k-1$ and $\sigma(x\alpha_h)$ is the same for all $x\in M_h$.
This gives $2k$ monochromatic copies of $K_{1,k-1}$ with centers $\alpha_h$ for $h\in [2k]$.
At least $k$ of them (say with centers $\alpha_h$ for $h\in [k]$) have the same color.
If this color is 0, these copies together with $\{u\alpha_h: h\in [k]\}$ and $F'$ give a monochromatic $J_k$ with color 0 and center $u$.
Otherwise, this color is 1.
Then these copies together with $\{x_j\alpha_h: h\in [k]\}$ and $F''$ give a monochromatic $J_k$ with color 1 and center $u$.
This proves $\Tr(100k)\to J_k$. \qed
\section{Monochromatic bistar}
In this section we prove $\Tr(6k+30)\to B_k$. We first establish the following lemma.
\begin{lem}\label{ll1}
Let $uvwu$ be the triangle bounding the outer face of $\Tr(k+10)$.
Let $\sigma: E(\Tr(k+10))\to \{0,1\}$ such that $|\{\sigma(ux): x\in V(\Tr(k+10))\}|=1$ and there is no monochromatic $C_4$ containing $u$.
Then $\Tr(k+10))$ contains a monochromatic $B_k$.
\end{lem}
\pf Without loss of generality, let $\sigma(uv)=0$. Let $x_0:=w$ and, for $i\in [6]$,
let $x_i\in V(\Tr(i))\setminus V(\Tr(i-1)$ such that $x_i$ is inside $uvx_{i-1}u$.
Since $\Tr(k+10)$ has no monochromatic $C_4$ containing $u$, we see that $|\{0\leq i\leq 6:\sigma(vx_i)=0\}|\le 1$.
So there exists some $i\in \{0,1,2,3,4\}$ such that
$\sigma(vx_i)=\sigma(vx_{i+1})=\sigma(vx_{i+2})=1$.
We have either $\sigma(x_ix_{i+1})=1$ or $\sigma(x_{i+1}x_{i+2})=1$;
as otherwise $ux_ix_{i+1}x_{i+2}u$ is a monochromatic $C_4$ of color 0 and containing $u$, a contradiction.
We consider two cases.
\medskip
\noindent {\it Case} 1. $\sigma(x_ix_{i+1})=\sigma(x_{i+1}x_{i+2})$.
In this case, we have $\sigma(x_ix_{i+1})=\sigma(x_{i+1}x_{i+2})=1$.
So $x_ix_{i+1}x_{i+2}vx_i$ is a monochromatic $C_4$ of color 1.
Let $y_0:=x_{i+1}$ and for $\ell\in [k+1]$, let $y_\ell\in
V(\Tr(i+1+\ell))\setminus V(\Tr(i+\ell))$ such that $y_\ell$ is inside $uy_{\ell-1}x_iu$.
Similarly let $z_0:=x_{i+1}$ and for $\ell\in [k+1]$, let $z_\ell\in V(\Tr(i+2+\ell))\setminus V(\Tr(i+1+\ell))$
such that $z_\ell$ is inside $uz_{\ell-1}x_{i+2}u$.
Since $\Tr(k+10)$ has no monochromatic $C_4$ containing $u$,
this shows that $|\{\ell\in [k+1]: \sigma(x_iy_\ell)=0\}|\le 1$ and $|\{\ell\in [k+1]:\sigma(x_{i+2}z_\ell)=0\}|\le 1$.
Therefore, there exist $Y\subseteq \{y_\ell: \ell\in [k+1]\}$ and $Z\subseteq \{z_\ell: \ell\in [k+1]\}$ such that
$|Y|=|Z|=k$, $\sigma(yx_i)=1$ for each $y\in Y$ and $\sigma(zx_{i+2})=1$ for each $z\in Z$.
Hence, $\Tr(k+10)$ has two monochromatic $K_{1,k}$s of color 1 with centers $x_i,x_{i+1}$ and leave sets $Y, Z$, respectively.
These two $K_{1,k}$s together with $vx_ix_{i+1}x_{i+2}v$ form a monochromatic $B_k$ of color 1.
\medskip
\noindent {\it Case} 2. $\sigma(x_ix_{i+1})\ne \sigma(x_{i+1}x_{i+2})$.
Without loss of generality, let $\sigma(x_ix_{i+1})=0$ and $\sigma(x_{i+1}x_{i+2})=1$.
Let $y\in V(\Tr(i+2))\setminus V(\Tr(i+1))$ be
inside $ux_ix_{i+1}u$. Because $\sigma(uy)=0$ and $\Tr(k+10)$ has no monochromatic $C_4$ containing $u$,
$\sigma(yx_i)=\sigma(yx_{i+1})=1$.
Therefore, $yx_{i+1}vx_iy$ is a monochromatic $C_4$ of color 1.
Let $y_0:=y$ and, for $\ell\in [k+1]$, let $y_\ell\in
V(\Tr(i+2+\ell))\setminus V(\Tr(i+1+\ell))$ such that $y_\ell$ is inside $uy_{\ell-1}x_iu$.
Let $z_0:=y$ and, for $\ell\in [k+1]$, let $z_\ell\in V(\Tr(i+2+\ell))\setminus V(\Tr(i+1+\ell))$
such that $z_\ell$ is inside $uz_{\ell-1}x_{i+1}u$.
The remaining proof is similar as in Case 1.
We observe that $|\{\ell\in [k+1]: \sigma(x_iy_\ell)=0\}|\le 1$ and $|\{\ell\in [k+1]:\sigma(x_{i+1}z_\ell)=0\}|\le 1$.
Therefore, there exist $Y\subseteq \{y_\ell: \ell\in [k+1]\}$ and $Z\subseteq \{z_\ell: \ell\in [k+1]\}$ such that
$|Y|=|Z|=k$, $\sigma(yx_i)=1$ for $y\in Y$, and $\sigma(zx_{i+1})=1$ for $z\in Z$.
Hence, $\Tr(k+10)$ has two monochromatic $K_{1,k}$s of color 1 with centers $x_i,x_{i+1}$ and leave sets $Y, Z$, respectively.
These two $K_{1,k}$s together with $yx_{i+1}vx_iy$ form a monochromatic $B_k$ of color 1.
This proves Lemma \ref{ll1}.
\qed
\medskip
We are ready to prove $\Tr(6k+30)\to B_k$.
Let $\sigma: E(\Tr(6k+30)\to \{0,1\}$.
By Lemma \ref{astu}, the copy of $\Tr(16)$ with the same outer face as of $\Tr(6k+30)$ contains a monochromatic $C_4$, say $u_1u_2u_3u_4u_1$ of color 1.
For each $i\in \{1,3\}$, let $v_iw_i$ be an edge in $\Tr(18)$ such that $u_iv_iw_iu_i$ is a triangle inside $u_1u_2u_3u_4u_1$.
Note that $u_iv_iw_iu_i$ bounds the outer face of a $\Tr(6k+12)$.
Let $A_i$ be the family of all vertices $x\in N(u_i)$ inside $u_iv_iw_iu_i$ and satisfying $\sigma(xu_i)=1$.
If $|A_1|\geq k$ and $|A_3|\geq k$, then together with the monochromatic 4-cycle $u_1u_2u_3u_4u_1$, it is easy to form a monochromatic $B_k$ of color 1.
Hence by symmetry, we may assume that $|A_1|<k$.
Then there exists an edge $vw$ in $\Tr(18+k)$ such that $u_1vwu_1$ bounds an inner face of $\Tr(18+k)$ and
$\sigma(u_1x)=0$ for all $x\in N(u_1)$ in the closed disc bounded by $u_1vwu_1$.
We may assume that the induced subgraph contained in the closed disc bounded by $u_1vwu_1$
has a monochromatic $C_4$ say $u_1xyzu_1$ (as otherwise, it contains a $B_k$ by Lemma~\ref{ll1}).
Furthermore, we have $\{x,y,z\}\subseteq V(\Tr(2k+28))$.
Let $\{p_0,q_0\}\subseteq V(\Tr(2k+29))\setminus V(\Tr(2k+28))$ such that both
$xyp_0x$ and $yzq_0y$ bound two inner faces of $\Tr(2k+29)$.
For $\ell\in [3k]$, let $p_\ell\in V(\Tr(2k+29+\ell))\setminus V(\Tr(2k+28+\ell))$ such that
$p_\ell$ is inside $xp_{\ell-1}yx$.
Similarly, for $\ell\in [3k]$, let $q_\ell\in V(\Tr(2k+29+\ell))\setminus V(\Tr(2k+28+\ell))$ such that $q_{\ell}$ is inside $yq_{\ell-1}zy$.
Moreover, let
$$B_1:=\{p\in N(x): p \mbox{ is inside } xp_0yx \mbox{ and } \sigma(xp)=0 \},$$
$$B_2:=\{q\in N(z): q \mbox{ is inside } yq_0zy \mbox{ and } \sigma(zq)=0\}.$$
If $|B_1|\geq k$ and $|B_2|\geq k$, we can find two monochromatic $K_{1,k}$s of color 0, one inside $xp_0yx$ rooted at $x$ and one inside
$yq_0zy$ rooted at $z$; these two $K_{1,k}$s and $u_1xyzu_1$ form a monochromatic $B_k$ of color 0.
So we may assume, without loss of generality, that $|B_1|<k$.
Let $C:=\{\ell\in [3k]: \sigma(yp_\ell)=0 \}$. We claim $|C|< k$.
Suppose on the contrary that $|C|\ge k$.
Then there is a monochromatic $K_{1,k}$ with root $y$ and leaves in $C$ of color 0.
Since $\sigma(u_1p)=0$ for all $p\in N(u_1)$ inside $u_1vw$,
there is also a monochromatic $K_{1,k}$ with root $u_1$ and leaves inside $u_1xyzu_1$ of color 0.
Now these two $K_{1,k}$s and $u_1xyzu_1$ form a monochromatic $B_k$ of color 0.
So $|B_1|<k$ and $|C|<k$. Then there exist $p_{h},p_{s}$ with $h,s\in [3k]$ such that
$\sigma(p_hx)=\sigma(p_hy)=\sigma(p_{s}x)=\sigma(p_{s}y)=1$.
Because $xp_0p_1x$ bounds an inner face of $\Tr(2k+30)$, it also bounds the outer face of a $\Tr(4k)$.
As $|B_1|<k$, there exists a monochromatic $K_{1,k}$ of color 1 with the root $x$ and $k$ leavers inside $xp_0p_1x$.
Similarly, as $|C|<k$, there exists a monochromatic $K_{1,k}$ of color 1 with root $y$ and $k$ leavers inside $yp_0p_1y$.
Now these two $K_{1,k}$s and the 4-cycle $xp_hyp_sx$ form a monochromatic $B_k$ of color 1.
This proves that $\Tr(6k+30)\to B_k$ and thus completes the proof of Theorem \ref{radius2}. \qed
%\section{Conclusion}
\begin{comment}
Finally, through Theorem \ref{main}, \ref{ax}, \ref{new} it is easy to know that if
$\Tr(n)\to H$ for some $n$ and the radius of $H=2$, then $H$ must be the
subgraph of $J_k$ or $B_k$ for some $k$. Moreover, By Theorem \ref{jelly},
\ref{bistar}, If H is a subgraph of $J_k$ or $B_k$ for some k, then
$\Tr(n)\to H$ for some n. As a result, we have determined all the graphs H
with radius 2 and $\Tr(n)\to H$ for some n.
\end{comment}
\begin{comment}
\bigskip
(2) Considering $Y_{(o)uvw}$, if CR(ov)=CR(ow)$\neq$ CR(ou), then all the other edges
with an end point o must be colored the same as ou.\\
We omit the proof since the result is a direct application of (1) and the coloring
method.\\
(3) ou is an edge of type2 in $Y_{(o)uvw}$. p is connected to both o and u and p is
not v or w. Then CR(pu)$\neq$CR(po)\\
We also omit the proof because it directly follows from the coloring method and (1)\\
(4) Suppose CR(ou)=0. If ou is of type3 in $Y_{(o)uvw}$, then all the red simple
paths between o and u are of length 1 or 2. Similarly, if CR(ou)=1, ou is of type3,
then all the green paths between o and u are of length 1 or 2.\\
Proof of (4):\\
We just prove the first part.\\
Suppose o, $a_{1}$,$a_{2}$...$a_{r}$,u is a simple red path. If r=0, then the path is
of length 1. Otherwise r$\geq$1. If $a_{1}$ in ovwo, then according to (2), all edges
except uv,vw,uw with an end point v or w are colored green. In that, $a_{1}$ cannot
reach u. Contradicts. If $a_{1}$ is in ouvo, then according to the coloring method,
there is a list of points $e_{1},e_{2},...e_{n}$ in ouvo, which together with points
o, u form a fish with red ribs and green spines(see figure 6). In fact, $e_{i+1}$ is
the center inner point of uo$e_{i}$u). Let v be $e_{0}$. Suppose $e_{n}$ appears in
the last iteration. Then either a1 belongs to \{$e_{1},e_{2},...e_{n}$\} or $a_{1}$
is in an o$e_{i+1}e_{i}$o($n-1\geq\i\geq0$). In the former case, according to (2),
r=1 since u and o are the only two vertices connected to $e_{i}$ by a red edge. In
the latter case, $a_{1}$ cannot reach u because all edges except $e_{i}u\ and\
e_{i}o$ connected to $e_{i}$ are green as a result of (2). Similarly, when $a_{1}$ is
in uowu, (4) also holds. In conclusion, (4) is true.\\
%\begin{figure}
% \centering
% \includegraphics[scale=0.4]{005.png}
% \caption{proof of (4)}
%\end{figure}\\
(5)Now we proof theorem1.1 by contradiction.\\
Assume \Tr(n) has a monochromatic $C_{k}$(k$\geq$5)($a_{1},a_{2},...a_{k}$). If
$C_{k}$ has an edge of type3, then k$\leq$3 according to (4). Contradicts. Therefore, all of $C_{k}$'s edges are of type1 and type2. If $C_{k}$ is red, then suppose
$a_{1}a_{2}$ is of type 2. Suppose $a_{2}$ is the center inner point of $ua_{1}wu$.
According to (2), w and $a_{1}$ are the only two points connected to $a_{2}$ by a red
edge. In that, $a_{3}$ is w. According to lemma (3), $a_{1}a_{3}$is of type 1 or type
3. If $a_{1}a_{3}$ is of type 3, then $a_{1}a_{3}$is red and k$\leq$4 because of (1)
the coloring method and (4). Contradicts. If $a_{1}a_{3}$ is of type 1, then
$a_{1}a_{3}$ is ab, ac or bc(see Figure 3). According to (1), $a_{1}a_{3}$ cannot be
ab or ac because CR($a_{1}a_{2}$)=CR($a_{2}a_{3}$). In that, $a_{1}a_{3}$ is bc.
However, that means CR($a_{1}a_{2}$)=CR($a_{2}a_{3}$)=1. Contradicts. If $C_{k}$ is
green,similarly, we know $a_{1}a_{3}$ can only be of type 1. Besides $a_{1}a_{3}$ is
bc(see Figure3). Notice that there is a list of points $e_{1},e_{2},...e_{n}$ which
together with b,c form a fish with red spine and green rib(see Figure 8) We can
prove that any green path from b to c is of length 1 or 2 just like the proof of (4).
In that, k $\leq$4. Contradicts. In conclusion, \Tr(n) doesn't have monochromatic
$C_{k}$(k$\geq$5).\\
%\begin{figure}
% \centering
% \includegraphics[scale=0.4]{006.png}
% \caption{proof of (5)}
%\end{figure}\\
\end{comment}
\begin{thebibliography}{99}
\addtolength{\baselineskip}{-1ex}
\bibitem {AA89} I. Algor and N. Alon. The star arboricity of graphs, {\it Discrete Math.} {\bf 75}
(1989) 11--22.
\bibitem {AH77} K. Appel and W. Haken, Every planar map is four colorable. I. Discharging, {\it Illinois J. Mathematics} {\bf 21} (3) (1977) 429--490.
\bibitem {AHK77} K. Appel, W. Haken, and J. Koch, Every planar map is four colorable. II. Reducibility, {\it Illinois J. Mathematics} {\bf 21} (3) (1977) 491--567.
\bibitem {ASTU19} M. Axenovich, U. Schade C. Thomassen, and T. Ueckerdt, Planar Ramsey graphs, {\it Electronic J. Combinatorics} {\bf 26} (4) (2019) \#P4.9.
\bibitem {Go05} D. Goncalves, Edge partition of planar graphs into two outerplanar graphs, in
STOC?ˉ05: Proceedings of the 37th Annual ACM Symposium on Theory of Computing, pages 504--512, 2005.
\bibitem {CCW86} L. J. Cowen, R. H. Cowen, and D. R. Woodall, Defective colorings of graphs in
surfaces: partitions into subgraphs of bounded valency, {\it J. Graph Theory}
{\bf 10} (1986) 187--195.
\bibitem {Ha65} S. Hakimi, On the degrees of the vertices of a directed graph, {\it J. Franklin
Institute} {\bf 279} (1865) 290--308.
\bibitem {HMS96} S. Hakimi, J. Mitchem, and E. Schmeichel, Star arboricity of graphs, {\it Discrete
Mathematics} {\bf 149} (1996) 93--98.
\bibitem {RSST98} N. Robertson, D. Sanders, P. Seymour, and R. Thomas, The four-colour theorem, {\it J. Combin. Theory Ser. B} {\bf 70} (1) (1997) 2--44.
\bibitem {Po90} K. S. Poh, On the linear vertex arboricity of a planar graph, {\it J. Graph
Theory} {\bf 14} (1990) 73--75.
\end{thebibliography}
\end{document}